Comparison of Analytical Methodologies Based on 1H and 31P NMR

Furthermore, a method comparison study was conducted and the agreement between NMR and conventional methods was evaluated by using the Bland and Altma...
0 downloads 19 Views 143KB Size
J. Agric. Food Chem. 2007, 55, 577−584

577

Comparison of Analytical Methodologies Based on 1H and 31P NMR Spectroscopy with Conventional Methods of Analysis for the Determination of Some Olive Oil Constituents PHOTIS DAIS,*,† APOSTOLOS SPYROS,† STELLA CHRISTOPHORIDOU,† EMMANUEL HATZAKIS,† GEORGIA FRAGAKI,† ALEXIA AGIOMYRGIANAKI,† EMMANUEL SALIVARAS,‡ GEORGE SIRAGAKIS,§ DESPINA DASKALAKI,| MARIA TASIOULA-MARGARI,| AND MANUEL BRENES⊥ NMR Laboratory, Department of Chemistry, University of Crete, P.O. Box 2208, Voutes, 71003 Heraklion, Crete, Greece, Omikron Laboratory, 95 Giannitson Str., Peristeri, 12131 Athens, Greece, Minerva Olive Oil Company, 31 Valaoritou Str. Metamorphosi, Attiki, Greece, Department of Chemistry, Section of Industrial and Food Chemistry, University of Ioannina, 45110 Ioannina, Greece, and Department of Food Biotechnology, Instituto de la Grasa, CSIC, Avenida Padre Garcia Tejero 4, 41012 Sevilla, Spain

The present study was designed to assess the agreement between analytical methodologies based on 1H and 31P NMR spectroscopy and conventional analytical methods (titration, gas chromatography, and high performance liquid chromatography) for measuring certain minor and major constituents (free acidity, fatty acids, iodine value, and phenolic compounds) of olive oil. The standard deviations of the NMR method were comparable to those of the conventional methods, except perhaps those of the total hydroxytyrosol and total tyrosol. Linear regression analyses showed strong correlations between NMR and conventional methods for free acidity, total hydroxytyrosol, total tyrosol, total diacylglycerols, (+)-pinoresinol, (+)-1-acetoxypinoresinol, and apigenin; good correlations for linoleic acid, free hydroxytyrosol, and free tyrosol; and weak correlations for oleic acid, linolenic acid, saturated fatty acids, and luteolin. Furthermore, a method comparison study was conducted and the agreement between NMR and conventional methods was evaluated by using the Bland and Altman statistical analysis. The distribution of the data points in the bias plot showed that 96.4% and 100% of the measurements of free acidity and iodine value, respectively, were within the limits of agreement of the two methods. For the remaining constituents of olive oil, the percentage of measurements, located within the limits of agreement, ranged from 94% to 98.5%. KEYWORDS: Olive oil; quantitative NMR; phenolic compounds; free acidity

INTRODUCTION

In the past decade, high-resolution NMR spectroscopy has emerged as a potential analytical tool for the analysis of vegetable oils and in particular olive oil (1-4). The amount of information in a NMR spectrum obtained in a fairly rapid manner combined with the easy sample preparation render this spectroscopic technique very attractive for the determination of the composition of olive oil. 1H NMR spectroscopy has provided information (1, 3-6) about lipid classes, fatty acid composition, unsaturation levels, and several minor compounds * Authortowhomcorrespondenceshouldbeaddressed.Tel: +302810545037. Fax: +302810545001. E-mail: [email protected]. † University of Crete. ‡ Omikron Laboratory. § Minerva Olive Oil Co. | University of Ioannina. ⊥ CSIC.

(sterols, squalene, terpenes, volatile compounds, and others), whereas 13C NMR, among others, gave unique information about the positional distribution of fatty acids on glycerol and the stereochemistry of unsaturation (1, 2). Recently, 31P NMR spectroscopy has been employed in olive oil analysis (7) supplementing 1H NMR and 13C NMR spectroscopy, especially in cases where strong signal overlap and dynamic range problems in 1H NMR spectra and/or long relaxation times of the insensitive 13C nuclei render the analysis of olive oil a difficult task (8). This methodology is based on the derivatization of the labile hydrogen atoms of olive oil constituents bearing hydroxyl and/or carboxyl groups with the phosphitylating reagent 2-chloro-4,4,5,5-tetramethyldioxaphospholane (1) according to the reaction scheme illustrated in Figure 1 and the use of the 31P chemical shifts to identify the labile centers (compound 2). Compound 1 reacts rapidly and

10.1021/jf061601y CCC: $37.00 © 2007 American Chemical Society Published on Web 01/09/2007

578

J. Agric. Food Chem., Vol. 55, No. 3, 2007

Figure 1. Reaction of hydroxyl and carboxyl groups of olive oil constituents with the phosphorus reagent 2-chloro-4,4,5,5-tetramethyldioxaphospholane (1).

quantitatively under mild conditions with the hydroxyl and/or carboxyl groups. This analytical approach has been used in the past (7-9) to detect and quantify minor constituents of olive oil, such as diacylglycerols, phenolic compounds, total free sterols, and free fatty acids (free acidity). Despite the fact that NMR spectroscopy plays an everincreasing role in the study of olive oil, comparison of the NMR results with those obtained by official and/or well-established (for phenolic compounds) analytical techniques is scarce in the literature and lacking any statistical analysis (6, 10-14). Such a comparison would be a rigorous and efficient validation of 1H NMR and in particular 31P NMR spectroscopy as a quantitative analytical method for the quality control and authentication of extra virgin olive oil. In the present study, we determined the amounts of certain constituents (fatty acids, iodine value, diacylglycerols, phenolic compounds, and free acidity) of a large number of olive oil samples by employing 1H NMR and 31P NMR spectroscopy, and we compared the NMR data with results obtained by independent laboratories using official and/or well-recognized methods of analysis. In particular, fatty acids concentration was determined by using 1H NMR spectroscopy and gas chromatography, free acidity by 31P NMR spectroscopy and titration, phenolic compounds by 31P NMR spectroscopy and high performance liquid chromatography, diacylglycerols by 31P NMR spectroscopy and gas chromatography, and iodine value by 1H NMR and titration. The statistical evaluation of the comparison was made according to Bland and Altman methodology (15). MATERIALS AND METHODS Olive Oil Samples. A total of 145 olive oil samples collected from various regions of Greece and extracted from different olive varieties (koroneiki, athinolia, tsunati, kolovi, andramitiani, and two local varieties from kerkyra and Pilion) and years of harvesting (2002-2003, 2003-2004, 2004-2005, and 2005-2006) were used in the present study. A total of 111 samples were extra virgin olive oils with free acidity e0.8% in oleic acid, whereas the remaining 34 samples were lampante olive oils characterized by much higher free acidity (.0.8% in oleic acid), especially those obtained from the Island of Kerkyra (Corfu). The choice of lampante olive oils was dictated by the need to widen the range of free acidity values in order to perform linear regression analysis, as will be shown later. Chemicals. NMR. Pinacol, phosphorus trichloride, and protonated solvents (analytical grade) for synthesis of the phosphorus reagent, pyridine solvent (99%), and deuterated chloroform were purchased from Sigma-Aldrich (Athens, Greece). GC. Analytical grade n-tetradecane, n-hexane, and the silylating agents trimethylchlorosilane and N,N-bis(trimethylsilyl)trifluoroacetamide, used for the analysis of diacylglycerols, were purchased from Labscan (Hasselt, Belgium). n-Triacodane, dilaurin, 1,3-diolein, 1,2diolein, and 1-palmito-3-stearoylglycerol were purchased from Sigma Chemicals Co. (St. Louis, MO). For the analysis of fatty acids, analytical grade methanol, heptane, and potassium hydroxide were purchased from Sigma Chemicals Co. (St. Louis, MO).

Dais et al. HPLC. Protocol A. HPLC-grade acetonitrile, methanol, 2-propanol, and water were purchased from Merck (Darmstadt, Germany). Methanol, acetonitrile, and hexane used for phenolic extraction were proanalysis grade and purchased from Merck (Darmstadt, Germany). The standards used for the identification and quantification of phenolic compounds were tyrosol, apigenin, and luteolin purchased from SigmaAldrich (Germany); vanillic acid, p-coumaric acid, and ferulic acid purchased from Merck-Schuchardt (Hohenbrunn, Germany); oleuropein7-glucoside purchased from Extrasynthese Co. (Genay, France), and (+)-pinoresinol purchased from Separation Research (Turku, Finland). Protocol B. Methanol, N,N-dimethylformamide, hexane, water, and phosphoric acid were all of HPLC grade purchased from Teknokroma, S. L. (Barcelona, Spain). Tyrosol was purchased from Sigma Chemicals Co. (St. Louis, MO). Apigenin and luteolin were obtained from Extrasynthe`se Co. (Genay, France). Titrations. All reagents, solvents (analytical grade), and standards for the determination of free acidity (diethyl ether, 95% ethanol, 0.1 M potassium hydroxide, phenolphthaleine indicator) and iodine number (sodium iodide, starch indicator, 0.1 M sodium thiosulfate, cyclohexeneacetic acid) were purchased from Sigma-Aldrich (Athens, Greece). Preparation of the Phosphorus Reagent. The derivatizing phosphorus reagent was synthesized from pinacol and phosphorus trichloride in the presence of triethylamine following the method described in the literature (16). However, to increase the yield of the reaction, we utilized n-hexane solvent and pyridine instead of benzene and triethylamine, respectively, suggested in the original method. This modification resulted in ∼45% yield of the product versus ∼20% obtained with the original method. Extraction of Phenolic Compounds. For the NMR study and protocol A of the HPLC analysis, phenolic compounds were extracted following the method developed by Montendoro et al. (17) using a mixture of methanol-water (80:20 v/v). The phenolic extracts were dissolved in methanol for injection or used immediately for sample preparation prior to 31P NMR measurements. For protocol B of the HPLC analysis, phenolic extracts of olive oils were obtained following the procedure described elsewhere (18), using N,N-dimethylformamide. Standards for HPLC Analyses. Protocol A. The standards used for quantification of phenolic compounds were oleuropein-7-glycoside for the quantification of hydroxytyrosol and hydroxytyrosol derivatives (dialdehydic form of elenolic acid linked to hydroxytyrosol, oleuropein aglycon); tyrosol for quantification of tyrosol and tyrosol derivatives (dialdehydic form of elenolic acid linked to tyrosol and the ligstroside aglycon); (+)-pinoresinol for quantification of the lignans (+)pinoresinol and (+)-1-acetoxypinoresinol; and, finally, luteolin and apigenin for quantification of the flavonoids luteolin and apigenin, respectively. Identification of phenolic compounds was achieved by comparing their retention time with those of standards, from UV absorption and from GC-MS analysis of hydroxytyrosol and tyrosol derivatives as described elsewhere (19, 20). Protocol B. Apart from hydroxytyrosol, tyrosol, luteolin, and apigenin, the remaining phenolic compounds hydroxytyrosol acetate, the dialdehydic form of elenolic acid linked to hydroxytyrosol, the dialdehydic form of elenolic acid linked to tyrosol, the oleuropein aglycon, and the ligstroside aglycon used as standards were obtained using a semipreparative 25 cm × 10 mm, i.d., 5 µm, Spherisorb ODS-2 (Teknokroma, Barcelona, Spain) HPLC column and a flow rate of 4 mL/min. The mobile phases and the gradient have been described previously (21). Individual phenols were quantified by a four-point regression curve on the basis of the standards obtained form commercial suppliers or from preparative HPLC as already described. HPLC Analysis of Phenolic Compounds. HPLC equipments and conditions used to quantify phenolic compounds in olive oil with protocols A and B have been described in refs 19, 20 and 18, 21, respectively. GC Analysis of Diacylglycerols and Fatty Acids. The AOACS official method (22) developed by Firestone et al. (23) for the determination of mono- and diacylglycerols by capillary gas chromatography was utilized for the present analysis. A Shimadzu-17A gas chromatograph (Shimadzu Co., Kyoto, Japan) was fitted with a flame ionization detector (FID) and a column injection system. Separation was carried out on a fused-silica J&W capillary column (15 m × 0.32

J. Agric. Food Chem., Vol. 55, No. 3, 2007

Methods for the Determination of Olive Oil Constituents mm id) (Agilent Technologies, Palo Alto, CA) coated with DB-5HT ((5% phenyl)-methylpolysiloxane) of 0.1 µm thickness. The European Union official method (24, 25) for the determination of saturated and unsaturated fatty acids by capillary gas chromatography was utilized for the present analysis. The preparation of the fatty acid methyl esters was achieved via trans-esterification with a cold methanolic solution of potassium hydroxide according to the International Olive Oil Council official method (26). A Carlo Erba HRGC 5300 gas chromatograph (Rodano, Italy) equipped with a flame ionization detector (FID) and a split injection system was used. Separation was performed on a fusedsilica capillary column (60 m, 0.25 mm id) (Thames Restek, Saunderton, UK) coated with Rtx-2330 (90% biscyanopropyl/10% phenylcyano propylpolysiloxane) of 0.2 µm thickness. The saturated fatty acids determined in the present study by GC were palmitic acid (16:0), heptadecanoic acid (C17:0), stearic acid (C18:0), arachidic acid (C20: 0), behenic acid (C22:0), and lignoceric acid (C24:0). Determination of Free Acidity and Iodine Value. These parameters were determined by employing the official methods of titration (24, 25). Sample Preparation for Spectroscopic Analysis. Sample preparation for the determination of olive oil constituents by employing 1H and 31P NMR experiments has been described in detail in previous publications (7-9). For the determination of free acidity and the diacylglycerol content, derivatization of the carboxyl and hydroxyl groups by the phosphorus reagent (1) was performed in olive oil without any posterior treatment (7, 8), whereas the polar extract was phosphitylated for the quantification of phenolic compounds (9). NMR Experiments. All NMR experiments were conducted on a Bruker AMX500 spectrometer operating at 500.1 and 202.2 MHz for proton and phosphorus-31 nuclei, respectively, at 30 ( 1 °C. Details of recording 1H and 31P NMR spectra and chemical shifts assignments can be found elsewhere (7-9, 27). Statistical Analysis. Basic statistics (mean, standard deviation, etc.), correlations between sets of data (linear regression), and scatter-plots (bias plots) were processed by using Statistica 7.1 for Windows (StatSoft Inc.). RESULTS AND DISCUSSION

The results of the analyses of the present olive oil samples by employing 1H and 31P NMR methodology and conventional analytical methods are summarized in three tables and are available as Supporting Information. These tables contain the results of free acidity for 137 olive oil samples determined by titration following the EC official method (24, 25) and by 31P NMR spectroscopy according to references cited (7, 9); phenolic compounds for 28 olive oil samples determined from the polar fraction of olive oils by HPLC using two experimental protocols (18-21) and by 31P NMR spectroscopy (9); 1,2- and 1,3diacylglycerols for 26 olive oil samples determined by GC following the AOACS official method (22) and by 31P NMR spectroscopy (7, 8); unsaturated and saturated fatty acid composition determined for 137 olive oil samples by GC according to the EC official method (24, 25) and 1H NMR spectroscopy (27); and the iodine value determined for 38 olive oil samples by the EC official titration method (24, 25) and 1H NMR spectroscopy (27). The composition of the unsaturated fatty acids (oleic acid, linoleic acid, and linolenic acid) and the saturated fatty acids (SFAs) in olive oils was calculated upon combining the various signal intensities in the 1H NMR spectra as shown elsewhere (27). Contrary to GC, the composition of each SFA could not be determined separately by 1H NMR spectroscopy (1, 5, 27). Instead, 1H NMR results reflect collectively the concentration of all SFAs in olive oils. Therefore, comparison of the 1H NMR and GC data for SFAs requires the summation of the individual SFA concentrations determined by the conventional GC method. The concentration of phenolic compounds for 28 extra virgin olive oils was determined by liquid-liquid extraction followed

579

by phosphitylation and recording the 31P NMR spectra of the phosphitylated compounds (9). The phosphitylated aromatic hydroxyl groups of tyrosol and hydroxytyrosol in their free and conjugated forms showed about the same chemical shifts in the 31P NMR spectrum. Therefore, the integrals of these two signals at δ 139.20 and 138.19 reflect the concentrations of total hydroxytyrosol and total tyrosol, respectively (9). On the other hand, HPLC was able to separate and quantify all forms of hydroxytyrosol and tyrosol (18-21). This means that the 31P NMR data for total hydroxytyrosol and tyrosol are to be compared with the summation of the individual concentrations of free hydroxytyrosol and tyrosol, hydroxytyrosol acetate, and the hydrolysis products of oleuropein and ligstroside, namely, the dialdehydic form of elenolic acid linked to hydroxytyrosol, the dialdehydic form of elenolic acid linked to tyrosol, the oleuropein aglycon, and the ligstroside aglycon as determined by HPLC. Both HPLC protocols A and B used in this study were able to determine phenolic compounds in olive oil. However, phenolic compounds extracted from 18 out of 28 extra virgin olive oil samples and quantified by employing the HPLC protocol B took advantage of the appropriate standards used for the quantification of hydroxytyrosol acetate and the hydrolysis products of oleuropein and ligstroside (see above). For the remaining 10 olive oils samples analyzed using the HPLC protocol A, hydroxytyrosol acetate was not quantified, because its concentration was negligible. Moreover, the quantification of the hydrolysis products of oleuropein and ligstroside using the HPLC protocol A was made with oleuropein and tyrosol standards, respectively, which have much different absorption coefficient than the aforementioned hydrolysis products (28). Thus, comparison of the 31P NMR and HPLC data for total hydroxytyrosol and total tyrosol was made by using data from 18 olive oil samples analyzed by the HPLC protocol B. Further inspection of the data of analysis reveals that the percentages of (+)-pinoresinol and (+)-1-acetoxypinoresinol determined by the HPLC protocol B were systematically about four times higher than those obtained with 31P NMR methodology, whereas comparable values were observed while using the HPLC protocol A. A possible explanation of this discrepancy could be the different liquid-liquid extraction procedures used for the HPLC analysis following protocols A and B. It appears that the use of N,N-dimethylformamide as an extraction solvent in protocol B resulted in a more efficient recovery of lignans (18). The NMR data were to be compared with those obtained by the conventional analytical methods. The normal plots of the differences between the NMR data and those of the conventional methods revealed one outlier in each of the data sets of total diacylglycerol, free hydroxytyrosol, apigenin, and luteolin that was removed in subsequent statistical analysis. One issue that should be discussed first is related to the repeatability of the measurements. Table 1 summarizes the repeatability of the NMR results and those obtained by conventional methods. The repeatability was calculated by means of the following equation according to ISO 5725 (29):

repeatability ) 2.8 × xSD2 SD2

(1)

is the variance of repeatability. The variance of repeatability for each compound in Table 1 was calculated from the standard deviation of a number (8-10) of interday measurements using the same olive oil sample and the same protocol of the corresponding analytical method. The data in Table 1 demonstrate that the repeatabilities associated with NMR and conventional measurements for most olive oil constituents are in general comparable, except perhaps those observed for total

580

J. Agric. Food Chem., Vol. 55, No. 3, 2007

Dais et al.

Table 1. Repeatability (r)a for Measurements Using NMR Spectroscopy and Conventional Analytical Methods NMR method

conventional methodb

0.07 0.13 3.72 0.56 0.28 0.14 0.28 1.71 10.79 14.47 11.01 1.22 2.19 0.32 3.44

0.06 0.11f 2.69 0.42 0.16 0.14 0.22d 6.55 2.18 123.00f 79.04f 7.06 10.14 0.59 2.88

free acidityc total diacylglycerolsc iodine valued oleic acidd linoleic acidd linolenic acidd SFAd hydroxytyrosolc,e tyrosolc,e total hydroxytyrosolc,e total tyrosolc,e (+)-pinoresinolc,e (+)-1-acetoxypinoresinolc,e apigeninc,e luteolinc,e

a Calculated from eq 1. b The conventional methods are mentioned in text. Determined by 31P NMR methodology. d Determined by 1H NMR methodology. e Includes extraction of phenolic compounds. f Calculated from pooled standard deviation. c

Table 2. Correlation Coefficient (R), Intercept (R), Regression Coefficient (β), 95% Confidence Interval for the Regression Coefficient, and Statistical Significance (p-value) of the Regression Analysis of the Dependent Variable (Determined by NMR Methodsa) to the Independent Variable (Determined by Conventional Methodsa)

(137)b

free acidity total diacylglycerols (25) iodine value (38) oleic acid (137) linoleic acid (137) linolenic acid (137) SFA (137) hydroxytyrosolc (27) tyrosolc (28) total hydroxytyrosold (18) total tyrosold (18) (+)-pinoresinold (18) (+)-pinoresinole (10) (+)-1-acetoxypinoresinold (18) (+)-1-acetoxypinoresinole (10) luteolinc (27) apigeninc (27)

R

a

β

95% CI of β

p-value

0.994 0.972 0.528 0.871 0.947 0.740 0.802 0.945 0.937 0.990 0.968 0.796 0.991 0.983 0.736 0.830 0.982

−0.056 0.046 40.933 3.374 0.631 0.260 3.790 3.664 5.708 −1.219 27.107 −0.898 −0.360 −1.860 7.379 0.555 0.385

0.978 0.966 0.481 0.963 0.926 0.602 0.719 0.896 0.877 1.010 0.958 0.259 0.829 0.283 0.660 0.661 0.921

0.960−0.996 0.865−1.067 0.219−0.743 0.871−1.055 0.873−0.980 0.508−0.695 0.780−1.007 0.811−1.080 0.795−1.009 0.929−1.090 0.827−1.090 0.155−0.363 0.737−0.920 0.255−0.311 0.165−1.156 0.478−0.844 0.847−0.995